Competition for Managers, Corporate Governance and Incentive Compensation

Transcription

1 Competition for Managers, Corporate Governance and Incentive Compensation Viral Acharya (NYU, CEPR, NBER) Marc Gabarro (Erasmus University) Paolo Volpin (LBS, CEPR) January 2012 Abstract We propose a model in which better governance incentivizes managers to perform better and thus saves on the cost of providing pay for performance. However, when managerial talent is scarce, firms competition to attract better managers reduces an individual firm s incentives to invest in corporate governance. In equilibrium, better managers end up at firms with weaker governance, and conversely, better-governed firms have lower-quality managers. Consistent with these implications, in a sample of US firms, we show that (i) better CEOs are matched to firms with weaker corporate governance and more so in industries with stronger competition for managers, and, (ii) corporate governance is more likely to change when there is CEO turnover, with governance weakening when the incoming CEO is better than the departing one. JEL classification: D82, G21, G18. Keywords: corporate governance, executive compensation, externalities. Authors addresses: vacharya@stern.nyu.edu; gabarro@ese.eur.nl; pvolpin@london.edu. Acknowledgments: We thank Yakov Amihud, Jennifer Arlen, Ramin Baghai, Lucien Bebchuk, Vidhi Chhaochharia, Jeff Coles, Martijn Cremers, Alex Edmans, Andrew Ellul, Julian Franks, Gerard Garvey, Steven Kaplan, Roberta Romano, Alan Schwartz, Henri Servaes, Michael Weisbach, participants at the 2010 AEA meetings in Atlanta, the 2010 Yale-ECGI-Oxford Conference, the Spring 2011 NBER Law and Economics, the 2011 NYU-Penn Law and Finance Symposium, the 2012 AFA meetings in Chicago, and in seminars at Boston College, DePaul, Emory, Georgia Tech, HEC Paris, London Business School, MIT and New York University. We are grateful for research support from the ESRC (Grant No. R ) and the London Business School s Centre for Corporate Governance.

2 1 Introduction The public outcry against the pay of investment bankers following the crisis of is just the latest manifestation of the ongoing debate on executive pay that has kept academics busy for the last twenty years. Executives receive large pay for performance when their firm does well and they are also paid well when their firm does poorly (for instance, in the form of severance payments and golden parachutes). Why are executives (and other professionals) paid so much and, apparently, independently of performance? The literature has evolved into two conflicting camps. The first one, starting with Jensen and Murphy (1990), argues that entrenchment, or poor corporate governance, allows managers to skim profits away from the firm in the form of high pay (see also Bertrand and Mullainathan, 2001, Bebchuk and Fried, 2004, among others). The second camp suggests an efficient explanation: better managers can generate greater value at larger firms and competition for scarce managerial talent forces large firms to pay managers a lot (see Rosen, 1981, and Gabaix and Landier, 2008). In this paper, we show that these views are not in conflict and there is in fact a natural link between them. We develop a model of the managerial labor market in which poor corporate governance and entrenchment arise because of competition in the market for managerial talent. Some firms on purpose choose lower governance and higher pay to attract and retain better managers. The key insight is that corporate governance affects the matching between managers and firms. Better governance may incentivize managers to perform better for a lower pay. However, it also reduces firms ability to attract the best managers. In our model, firms can incentivize managers to choose the right action via (i) pay for performance, that is, by rewarding them when things go well, and (ii) corporate governance, that is, by punishing them when things go badly. When firms do not have to compete with each other to attract top quality managers, they choose an efficient combination of pay for performance and corporate governance that just meets the manager s incentive compatibility condition. 1

3 However, when managerial talent is scarce and firms have to compete to attract the few top quality managers, firms depart from the optimal level of corporate governance. This result follows from the inability of a firm to affect the rents of the top quality managers as these managers can always find another firm to employ them. In other words, the individual rationality constraint is binding and thus the overall compensation of top-quality managers is exogenous for a given firm. Therefore, it becomes inefficient for a firm that wants to employ a top quality manager to set high levels of corporate governance as it would in any case have to match the manager s individual rationality constraint by setting a high pay for performance. Thus, shareholders would end up bearing the full costs of better corporate governance in the form of higher executive pay, while they would share (for instance, with potential raiders) the benefits of corporate governance (in the form of more takeovers). With ex-ante identical firms and observable managerial talent, in equilibrium firms are indifferent between hiring a better- and a worse-quality manager. Then, the better-quality managers extract all the rents, which are exactly equal to the difference in profitability between better and worse managers. The firms that hire worse-quality managers feature the optimal combination of corporate governance and managerial pay. Those that hire better-quality managers rationally choose to underinvest in corporate governance and pay managers more. The main result of the model is that, in equilibrium, some firms attract better managers by paying them more and choosing more lax governance standards; others attract weaker managers by paying them less and choosing stricter governance standards. These associations are ex-ante rational as firms offered these compensation and governance packages to attract scarce managerial talent. If we can measure managerial talent, our main empirical prediction is that better quality managers are matched to firms that have weaker governance and receive higher pay. Moreover, takeovers should be negatively correlated with CEO quality, as takeover defenses are a principal form of weaker governance that can be offered to attract better-quality managers. Finally, since governance is part of optimal compensation package, changes in corporate governance should primarily arise when there is CEO turnover and should depend on the quality of the new CEO relative to the old 2

4 one: governance standards should improve when the new CEO is of worse quality than the old one and should worsen when on the contrary the new CEO is of better quality than the old one. We test these predictions using a dataset that combines balance-sheet data from Compustat on unregulated firms in the United States over the period 1993 to 2007, data from ExecuComp on the compensation they award their CEOs and on their turnover, M&A data from Thompson Deals, and firm-level corporate governance data from Riskmetrics. We focus on the G-Index developed by Gompers et al. (2003) and its individual components as our measures of outside corporate governance and find evidence in favor of our predictions. To show that the allocation of CEOs and firms is consistent with the matching equilibrium predicted by the model, our cross-sectional test follows a two-stage approach. In the first stage, managerial talent is measured as the CEO fixed effect in a regression of firm s operating performance on several control variables. In particular, we extract a measure of the CEO s talent relative to other CEOs in the industry. In the second stage, we correlate these predicted measures of managerial talent with corporate governance, executive compensation, firm size, CEO tenure and takeovers. We find that better managers are employed by larger firms, face weaker governance regimes, are paid more, are less likely to be replaced, and are less likely to be taken over, results that are consistent with the model s predictions. We find these associations even after controlling for proxies of CEO power (his tenure, age and whether he is externally hired), thus alleviating the concern that CEO power is the omitted variable behind the association between greater CEO quality and weaker firm governance. Moreover, we show that there is a stronger negative relationship between corporate governance and CEO quality in industries with greater competition for managers, as measured by the frequency of external hires. In time-series tests, we show that the changes in governance primarily happen around CEO turnovers. Further, as predicted by the model, when the new CEO is better than the old one, the quality of corporate governance decreases; while governance increases if the new CEO is of worse quality than the old one. When we examine which components of the G-Index are more correlated with changes in CEO 3

5 talent we find that the most important provisions are the ones shielding directors and officers from legal liability and the ones restricting shareholder voting rights. These provisions increase when the new CEOs are better than the old ones, both in a cross-sectional setup when we compare different firms and in a time-series setting when we consider CEO turnover. These provisions empower the CEO and enable him to fight takeovers more effectively. The results for the other sub-indexes are not robust across specifications. In particular, we do not find any significant relationship between State Law and CEO quality. This latter finding alleviates concerns that the G-Index is measuring anti-takeover rules outside the control of individual firms and indicates that our findings seem to come from differences in corporate governance arrangement that are indeed within the control of shareholders. The evidence from these tests provides support for our theoretical starting point that competition amongst firms for scarce managerial talent is an important determinant of observed executive compensation and governance practices. The rest of the paper is structured as follows. Section 2 discusses related literature. Section 3 presents the model. Section 4 presents the empirical evidence for our testable hypotheses. Section 5 discusses robustness checks and alternative explanations. Section 6 concludes. 2 Related Literature This paper is related to a large literature on executive compensation and corporate governance, but our approach is closer to the structural modeling of endogenous corporate choices as analyzed most recently in Coles, Lemmon and Meschke (2011), and the references therein. The neoclassical view is that executive compensation is the solution of the principal-agent problem between a set of risk-neutral investors and a risk-averse manager (Holmström, 1979). In this setting, pay for performance solves the trade-off between the need to incentivize the manager and the desire to insure him against idiosyncratic risk. According to this view, a firm chooses low- or high-powered compensation packages depending on the relative importance of managerial risk-aversion 4

6 and incentives. Starting with Jensen and Murphy (1990), skepticism grew among academics on whether this view provides a satisfactory explanation for the recent trends in executive compensation. Two alternative economic views have been suggested to explain executive compensation trends: one, managerial rent extraction, and second, efficient matching between managerial skills and firm characteristics. The first explanation links executive compensation to managers ability to extract rents (see Bertrand and Mullainathan, 2001, Bebchuk and Fried, 2004, Kuhnen and Zwiebel, 2009). According to this view, weaker corporate governance allows managers to skim profits from the firm, thereby leading to higher executive compensation. Even though this is currently the most popular explanation for the high executive pay, it begs several questions: If better corporate governance is the solution to excessive executive compensation, why don t all shareholders demand better corporate governance? Moreover, why are CEOs of well-governed firms also paid a lot? In our model, we treat corporate governance as a choice of the firm. We show that better corporate governance could indeed reduce managerial pay. However, when there is an active market for scarce managerial talent, firms are forced to choose weaker corporate governance and to leave rents for managers. In this respect, our contribution is to clarify the link between corporate governance, pay for performance and scarcity of managerial talent. The second explanation relates the level of executive pay to exogenous heterogeneity in firm size. Gabaix and Landier (2008), Terviö (2008), and Edmans, Gabaix and Landier (2009) present matching models à la Rosen (1981) in which the differences in size across firms predict some of the well-documented empirical facts on executive compensation. Gabaix and Landier (2008) and Terviö (2008) show that the empirically documented positive cross-sectional correlation between firm size and compensation may optimally arise in a setup where managerial talent has a multiplicative effect on firm performance and managers are compensated according to their increase in productivity as better managers will be matched to larger firms. Similarly, Edmans, Gabaix and Landier (2009) present a model in which both low ownership concentration and its negative correlation with firm size arise as part of an optimal 5

7 contract. 1 In a similar setup, Edmans and Gabaix (2011) show that inefficient incentive contracts and CEO allocation across firms arise when firms differ in terms of risks or disutilities for managers. Our model departs from this part of the literature because we treat firm size as an endogenous variable. In particular, we explore the impact of the extent of real investment on the market for managerial talent and corporate governance. We show that investment size may be a viable way to attract better managers and thereby determine the equilibrium choice of size by ex-ante identical firms. We find that indeed firms that invest more will attract better managers but will choose worse corporate governance. Conversely, firms that invest less will attract worse managers and will choose better corporate governance. Inefficiently low choices of governance and over-investment emerge as equilibrium outcomes because of the externality associated with the competition for managerial talent. Also, managers in our model can be incentivized by shareholders through a combination of incentive contracts and corporate governance, where governance acts as a substitute for compensation, as shown by Core et al. (1999) and Fahlenbrach (2009). Fahlenbrach (2009), in particular, finds that there is more pay for performance in firms with weaker corporate governance, as measured by less board independence, more CEO-Chairman duality, longer CEO tenure, and less ownership by institutions. Similarly, Chung (2008) studies the adoption of the Sarbanes-Oxley Act of 2002 and shows that firms required to have more than 50% of outside directors (interpreted as an improvement in shareholder governance) decreased significantly their CEO payperformance sensitivity relative to the control group. Finally, this paper is also related to a growing literature on spillover and externality effects in corporate governance initiated by Hermalin and Weisbach (2006), 1 Within this framework, the recent rise in compensation can be related to changes in the types of managerial skills required by firms. Murphy and Zábojník (2007) argue that CEO pay has risen because of the increasing importance of general managerial skills relative to firm-specific abilities. Supportive evidence is provided by Frydman and Saks (2010). Cremers and Grinstein (2010) study CEOs movements for the period between 1993 and 2005 and find that the characteristics of the market for CEOs differs across industries. Specifically, the proportion of CEOs coming from firms in other sectors significantly varies across industries, indicating that there is not a unique pool of managers that all firms compete for, but instead many pools specific to individual industries. 6

8 who provide a framework for assessing corporate governance reforms from a contracting standpoint and justify the need for regulation in the presence of negative externalities arising from governance failures. Acharya and Volpin (2010) and Dicks (2010) formalize this argument in a model where the choice of corporate governance in one firm is a strategic substitute for corporate governance in another firm. As in this paper, the externality therein is due to competition for managerial talent among firms. In a somewhat different context, Nielsen (2006) and Cheng (2011) model the negative externalities caused by earnings manipulation across firms. Nielsen (2006) considers a setting where governance improves publicly disclosed information about a firm and facilitate managerial assessment in competing firms. Cheng (2011) shows that earnings management in one firm may induce earnings management in other firms in the presence of relative performance compensation. 3 Theoretical Analysis The basic idea of our model is that firms compete for managers by choosing governance as part of an optimal incentive contract. We show in this section that in the presence of competition for scarce managerial talent, in equilibrium, ex-ante identical firms are indifferent between hiring a better manager, investing more and choosing weaker governance regime, and hiring a worse manager, investing less and setting a stronger governance regime. 3.1 Setup of the Model Consider an economy with n firms and m managers. There are two types of managers, m H are high-quality, well-established managers with a strong track-record (H-type), and m L are low-quality, or less-experienced, managers (L-type). All type L have low productivity e L < 1. H type manager s productivity is uniformly distributed over the [e L, 1] interval. Productivity is observable: we consider the case of unobservable productivity in the extension. We assume that the number of L-type managers is greater than the number of firms, m L > n, while the H-type managers are not numerous enough to be hired by all firms, m H < n. Managers and shareholders are 7

9 risk neutral. All firms are ex-ante identical. The timeline is as described in Figure 1: At t = 1, each firm s founder hires a CEO from a pool of candidates of observable quality e [e L, 1]. Given that abilities are observable, each firm sets a compensation contract which is a function of the manager s quality e. Managers apply for one of the jobs. If a manager is not employed at the end of this stage, he receives a reservation utility equal to 0. Similarly, a firm that does not employ any managers receives an output equal to 0. Compensation contracts are represented by a performance-related bonus p 0, which is contingent on the verifiable output X produced at t = 4. 2 Moreover, as part of the incentive package, at t = 1 the firm also chooses the investment size I 0 at a cost ri (with r > 1 being the cost of capital) and the level of corporate governance g [0, 1]. As we explain below, the benefit of corporate governance is that it increases the probability of a takeover at t = 3, and thus reduces managerial entrenchment. At t = 2, managers choose action Z {M, S}, where choice M generates a private benefit B for the manager and no output (X = 0) for the firm; while action S generates output X = Y (I) > I with probability ẽ and X = 0 otherwise, and no private benefit for the manager. We assume Y (I) > I, Y > 0, Y < 0, lim I 0 Y (I) =, lim I Y (I) < 1 to ensure an internal solution for the choice of investment. The choice of action is not observable by shareholders and the manager must stay employed until t = 4 for the firm to produce output X. At t = 3, shareholders and managers observe a signal x {Y (I), 0} on the expected output X. After observing this signal, shareholders can sell to a raider. The value produced by the raider, Y R, is uniformly distributed over the interval [0, R], with R < Y (I). With probability g, the manager has no power to stop the takeover: the takeover happens if the raider and the target shareholders agree on a price. With probability 1 g, the incumbent manager can fight the takeover and thus needs to agree for the takeover to succeed. 3 2 This assumption is without loss of generality because allowing for a further payment that is independent of performance would be inefficient: it would simply increase the amount of compensation needed in the case of good performance. 3 We assume that the firm must always settle any promised pay to the manager due at t = 4 even if the takeover happens and independently on whether the takeover requires the manager agreement 8

10 At t = 4, output is realized and distributed, the performance-related bonus p is paid, and, if still in control, the initial manager receives a private benefit b < R. For simplicity, we assume that (i) the manager has no negotiation power and thus only needs to be compensated for his outside option; and (ii) target shareholders receive a fraction δ [0, 1] of their joint surplus with the raider (net of any compensation for the incumbent manager). We make the following technical assumptions: (1) e L Y (I) B + (1 e L ) δr ri 0, with the condition being met with quality 2 for I = I L : profits from hiring the low ability manager are normalized to zero. (2) Y (I) B ri b > 0 holds for some value of I: hiring the highest ability manager, e = 1, is profitable even without considering the raider. (3) The signal x is perfectly informative. This assumption can be relaxed without substantially changing the model. (4) When indifferent, firms prefer to hire a H-type manager rather than a L-type one: this tie-breaking assumption simplifies the analysis. 3.2 Competition for Managers To derive the equilibrium, we proceed by backwards induction, starting from the takeover at t = Takeover If x = Y (I), there is no takeover opportunity as no raider can produce an output greater than Y (I). If instead x = 0, then the raider can always increase productivity. With probability g, the incumbent manager has no control on the takeover process and thus does not need to be compensated for the loss of private benefits. Thus, the surplus that the raider can share with target shareholders is Y R and target shareholders are paid δy R. The takeover happens with probability 1. to succeed. This requirement ensures that firms do not behave strategically and fire a manager that will produce higher output purely to save on pay for performance. 9

11 With probability 1 g instead, the incumbent manager must be compensated for the loss of private benefits. Hence, he is paid b and the surplus from the takeover that the raider and target shareholders share is Y R b. The takeover probability is thus equal to the probability that Y R > b or 1 (b/r). Therefore, the expected takeover price equals δr/2 if the manager cannot fight the takeover (which happens with probability g) and δ R (R b) 1 (R b)2 dr = δ if b R 2R the manager can fight the takeover (which happens with probability 1 g). Hence, the expected payoff for target shareholders if x = 0 is P = g δr 2 + (1 g)δ(r b)2 2R = δ [ (R b) 2 + gb(2r b) ]. (1) 2R Notice that P is strictly increasing in corporate governance g: the benefit of high corporate governance is a higher expected value of the firm in the bad state of the world x = Moral Hazard Problem Now consider the manager of type e s incentive compatibility and participation constraint at t = 2. Starting with the incentive compatibility condition, if the manager chooses the private-benefit action Z = M, output always equals 0 and manager s utility equals: U(M) = B + (1 g) b (2) The first term is the private benefit B from choosing action M, while the second term is the private benefit b from staying in control, which is paid only with probability 1 g. If he chooses the firm-value maximizing action Z = S, then his utility equals U(S) = ep + (1 e) (1 g) b (3) Hence, we can derive the incentive compatibility (IC) condition U(S) U(M) as p B e + (1 g) b. (4) Provided that the IC constraint is satisfied, the corresponding individual rationality (IR) constraint becomes p u e (1 e) (1 g) b e (5) 10

12 where u e is type e s reservation utility and will be endogenously determined so as to clear the market for managers. These constraints highlight the role of corporate governance from the manager perspective. Increasing corporate governance implies that the incentive compatibility condition is achievable with lower compensation. However, the same decrease in corporate governance implies that higher pay is necessary to meet the individual rationality condition. This is because increasing corporate governance increases the probability of a non-managerial agreed takeover, increasing the probability of losing b Incentive Contract Proceeding backwards to t = 1, shareholders expected profits equal Y (I) p if the project is successful (which happens with probability e) and the expected takeover price given by equation (1), if the project fails (which happens with probability 1 e). Since the net of investment cost is ri, shareholders problem is: δ [ max e [Y (I) p] + (1 e) (R b) 2 + gb(2r b) ] ri (6) (p,g,i) 2R subject to the IC and IR conditions (4) and (5). Analyzing the optimal incentive contracts conditional on the manager s type, in Appendix A we derive the following result: Lemma 1: The optimal investment for a firm hiring a manager of type e is I e = Y 1 ( r e). The optimal contract for a manager of type e depends on ue : (i) if u e < B, the optimal incentive contract is (p e, g e ) = ( B e, 1), with associated profit equal to ey (I e ) B (1 e) δr 2 ri e; (ii) if u e [B, b + B], the optimal incentive contract is (p e, g e ) = ), 1 ue B, with associated profit equal to ey (I e b e ) u e + (1 ( ue (1 e)(u e B) e) 1 2R {δr2 + (u e B) [2R (1 δ) + δb]} ri e ; and (iii) if u e > b + B, the optimal incentive contract is (p e, g e ) = ] associated profit equal to ey (I e ) u e + (1 e) [b + δ(r b)2 ri 2R e. ( ) ue (1 e)b, 0 with e Intuitively, the optimal incentive contract for a type-e manager depends on his reservation utility. If the manager reservation utility is very low (u e < B), the 11

13 individual rationality constraint is redundant and thus firms can extract all possible surplus from the manager by increasing corporate governance to the maximum level and saving on pay. If instead manager s reservation utility is large (u e > b + B), governance does not work as a substitute for pay. Hence, shareholders fully internalize the costs of the takeover borne by the manager, and understand that compensating the manager with poor governance is efficient (compared with paying higher incentive pay). The intermediate case in which u e [B, b + B] is the most interesting. Figure 2 shows the IC and IR constraints in the (g, p) space. The IC constraint is the downward sloping line with intercept b + B/e while the IR constraint is the upward sloping line with intercept [u e (1 e)b] /e. The set of feasible choices of (g, p) is the shaded area in Figure 2. Given that shareholders objective function is increasing in g and decreasing in p, the solution must lie on the IR constraint on the right of the IC line. Because the shareholder profits increase in g slower than managerial utility decreases in g, the optimal choice is the minimum level of corporate governance which satisfied both constraints. Hence, the optimal contract is at the intersection of the two lines (point A in the figure), where g e = 1 ue B b and p e = ue (1 e)(ue B) e. As proved in Appendix A, the equilibrium in the managerial market is a follows: Proposition 1 (Competition for managerial talent) A mass m H of firms hire a H manager. The compensation contract for an H-type manager with ability e is ( u 1 e (1 e)b, 0, Y ( 1 r e e) ) if e > ê ( (p e, g e, I e ) = u 2 e (1 e)(u2 B) e, 1 u2 e B, Y ( 1 r e b e) ) if e ê where ê, u 1 e and u 2 e are defined in Appendix A. The remaining n m H firms hire L type managers and offer the contract ( ( )) B r (p L, g L, I L ) =, 1, Y 1. e L This is the key result of the model. Because there is a scarcity of H-type managers, in equilibrium, competition among firms will be so that the rent awarded to H- type managers (u e ) makes firms indifferent between hiring a H-type or a L-type e L 12

14 manager. If hiring a H-type manager leads to higher profits than hiring a L-type manager, then a firm can marginally increase the compensation to H-type types, attracting one of them for sure, increasing profit. If instead hiring a L-type manager leads to a higher profit, all firms would hire a L-type manager and thus H-type managers would be willing to work for less. Given that corporate governance is used by firms to reduce managerial rent, firms hiring H-type managers find high level of corporate governance suboptimal. Conversely, firms hiring L-type managers face no competition for them and can, therefore, keep managerial compensation down to the incentive compatibility constraint. Thus, these firms choose the profit-maximizing level of corporate governance. The solution also highlights a potential reason for the non-perfect substitutability of corporate governance and executive compensation. Proposition 1 shows that firms mechanism to increase rents to the H-type managers is to choose a suboptimal level of corporate governance instead of implementing the optimal level of corporate governance and increasing executive compensation. The reason is because shareholders do not internalize the externality their choices of corporate governance impose on other firms stakeholders. Specifically, in our model, when firms increase corporate governance, they increase the probability of takeover, but firms only obtain a fraction δ of this increase in takeover profits. However, given that the individual rationality condition is binding, the firm must compensate the manager for the entire cost of this increase in corporate governance. Hence, the firm bears all the cost of higher governance but only enjoys part of the benefits. 3.3 Extension: No Competition for Managerial Talent We have assumed so far that managerial quality is perfectly observable. This is an important assumption but it can be relaxed. The results can be extended to the cases in which there are only imperfect signals about the quality of managers. As long as these signals contain some information, so that managers have different expected utility, the analysis would follow similarly. If instead, there is no information about the quality of managers, the results are quite different. We define e as the expected ability of a randomly picked manager. In 13

15 that case, since all managers are ex-ante identical and they are more than the number of firms (m H + m L > n), there is no effective competition for managers. Hence, the manager s outside option is equal across types and equal to the reservation utility from being unemployed (u e = 0). Therefore, from Lemma 1, Proposition 2 (No competition for managerial talent) All firms offer the contract and hire any manager. (p e, g e, I e ) = ( B ( r ) ) e, 1, Y 1, e Notice that the choice of corporate governance is (on average) higher than in the case with known type and competition among firms for scarce managerial talent. This results emphasizes the main intuition behind the model: it is the competition for scarce managerial talent (and not the higher ability of some managers) that affects the choice of corporate governance. Specifically, we can compare the contract that a manager with ability e > e L is offered under the two informational contexts. In the { } case with known managerial talent g e = max 1 u2 e B, 0 < 1, while in the case b with unknown managerial talent g e = 1, a higher level of governance than in the case with known managerial talent. 4 Empirical Analysis We now turn to the empirical analysis. First, we develop the two main testable implications of the model. Then, we present the empirical methodology. Finally, we describe our data and discuss our results. 4.1 Empirical Predictions The main result of the model is that in equilibrium some firms will attract better managers by paying them more, choosing weaker governance standards and larger 14

16 size; others will attract worse managers by paying them less, choosing stricter corporate standards and smaller size. 4 Thus, provided that we can find an appropriate measure of managerial talent, our main empirical prediction is: Prediction 1 (Matching equilibrium): Better quality managers receive higher pay, are matched to firms that have weaker governance standards, larger size and are less likely to be taken over. The comparison between Propositions 1 and 2 highlights how the role of corporate governance as part of an optimal compensation contract depends on the degree of competition for managers. Specifically, our model predicts that better managers are matched to firms that have lower corporate governance only when the competition among firms to attract them is high. Therefore, conditionally on us finding a relevant measure of the effective competition for managers, our model predicts: Prediction 2 (Competition for Managers): The more intense the competition for managers, the greater is the role of corporate governance as part of an optimal compensation contract. Specifically, the relationship between high quality managers and weaker governance standards should be stronger in sectors with stronger competition for managers. 4.2 Empirical methodology To test our main empirical prediction, that is, Prediction 1, we need to develop a measure of managerial ability (γ j ). For this purpose, we follow Bertrand and Schoar (2003) and Graham, Li and Qiu (2008) and compute the (unobserved) CEO impact on performance, where the latter is measured by return on assets. 5 The idea is to attribute to CEO ability the return on assets in excess of the value predicted by e ê. 4 Given that ūe e > 0, it is straight forward to prove that g e 0, with a strict inequality for 5 An alternative approach is to proxy CEO quality with observable characteristics, like press coverage (as in Milbourn, 2003) or MBA education (Murphy and Zaboknik, 2007). 15

17 firm-level and time-varying control variables. More precisely, we estimate ROA it = βx it + δ t + z ind + γ j + ε it, (7) where ROA it stands for return on assets for firm i in period t. X it are some time variant firm characteristics, which include size, book leverage, cash, interest coverage, dividend earnings, Tobin s q and governance measures; δ t are time fixed effects; and z ind are industry fixed effects. The parameter γ j is a fixed effect for a CEO, i.e., a dummy variable that takes value one when CEO j works in firm i and zero otherwise. This is our measure of managerial ability as it captures the unobserved (and time invariant) managerial effect on return on assets relative to the industry. Thus, the crucial identification strategy for our model is that the firm could have attracted any other manager in their industry if it wanted. Cremers and Grinstein (2010) document that most of the managerial mobility takes place within an industry so industry dummies constitute a natural starting point. 6 We test Prediction 1 in two ways: (i) by focusing on the cross-sectional differences across firms; and (ii) by emphasizing the time-series changes within a firm Prediction 1: Cross-sectional test In the first approach, we use the estimated fixed effects γ j as regressors in the following specification: Y it = β γ γ j + υ jt + χ t + z ind + ξ it, (8) where γ j are the CEO Ability coefficients estimated from regression (7); χ t and z ind are time and industry dummies and υ it are a set of CEO characteristics. Time dummies should control for any time pattern while industry dummies control for the average quality of CEOs hired in a given industry. We correct for the fact that γ j are generated regressors by weighting each observation by the inverse of the γ j standard error from the first-stage estimation. 6 To control for any endogenous manager-firm matching, we repeat the estimation of γ j including firm fixed effects instead of industry dummies. Results are qualitatively similar but some coefficients lose statistical significance. Including only industry dummies is more appropriate to preserve the power of the test, given the low mobility of CEOs across firms. 16

18 We estimate the specification above for different dependent variables Y it, that correspond to different empirical predictions. Y it will in turn be our measures of corporate governance, executive compensation and firm size. Our model predicts that (i) better managers work in firms that have lower corporate governance (that is, we expect β γ < 0 when Y it =Governance it ); (ii) better managers are paid more (that is, β γ > 0 when Y it =Compensation it ); and (iii) better managers work in larger firms (that is, β γ > 0 when Y it =Firm Size it ). An additional empirical implication of our model is that poor-quality managers should face stronger corporate governance, for instance, takeovers, more often. Therefore, we also estimate the specification above with Y it being a dummy variable for takeovers, with our model predicting β γ < 0. Moreover, takeovers are only one of the many mechanisms available to replace managers. In our model when a firm expected output is low, managers get replaced. As lower quality managers obtain low outputs more frequently, they are replaced more frequently. We test this empirical prediction using a duration model as follows h[t; X(t)] = F (β γ γ j + υ jt + χ t + z ind + ξ it ) (9) where h[t; X(t)] is the hazard function; defining the failure event as manager turnover. As usual, h[t; X(t)] describes the instantaneous rate of turnover at T given that there has been no turnover until t. As above, χ t and z ind are time and industry dummies and υ it are a set of CEO characteristics. The model s prediction is a positive correlation between CEO quality and employment length. To sum up, we test the main prediction of the model by running a within-industry two-stage analysis. In the first stage, we obtain individual CEO skills relative to the other CEOs employed in the industry from specification (7). In the second stage, we run regressions (8) and (9), to test whether these relative CEO abilities (compared with other CEO abilities in the industry) are correlated with corporate governance, CEO compensation, firm size and turnover, as predicted by our model Prediction 1: Time-series test Our model highlights the role of corporate governance as part of an optimal compensation contract. Therefore, changes in corporate governance should happen around 17

19 turnover, when the new compensation contract is agreed. More precisely, we should observe a negative correlation between the change in the manager quality and the change in firms governance standards. To test this prediction we estimate the following logit model: Governance Chg it = F (β T T urnover it + υ jt + χ t + z ind + ξ it ), (10) where Governance Chg it measure the changes in corporate governance and Turnover it captures the changes in managerial ability ( γ j ). As in all the previous regressions, χ t and z ind are time and industry dummies and υ it are a set of CEO characteristics Prediction 2 Finally, we test whether the use of corporate governance as part of an optimal compensation contract varies across sectors as a function of the degree of the competition for managers in those sectors (Prediction 2). To do so, we repeat the estimation of equation (8) separately for each Fama-French 49 industry, Governance it = β ind γ γ j + υ jt + χ t + ξ it (11) where γ j are the CEO Ability coefficients estimated from regression (7), χ t are time dummies, υ it are a set of CEO characteristics and ind is a different code for each industry. Then, we estimate the correlation between the different coefficients β ind γ our measure of the competition for managers: the percentage of insider promotions calculated by Cremers and Grinstein (2010). The identification assumption is that sectors with more frictions to cross-firm mobility would be associated with both a larger number of internal promotions and a lower degree of competition for managers. Therefore, our model predicts a negative correlation between β ind γ of insider promotions. and and the percentage 4.3 Data description We use firm-level financial variables from Compustat: ROA is the ratio of EBITDA (item ib) over lagged total assets (item at); Cash is cash and short-term investments (item che) over net property, plant, and equipment at the beginning of the fiscal 18

20 year (item ppent); Interest Coverage is earnings before depreciation, interest, and tax (item oibdp) over interest expenses (item xint); and Dividend Earnings is the ratio of the sum of common dividends and preferred dividends (items dvc and dvp) over earnings before depreciation, interest, and tax (item oibdp). We define Book Leverage as the ratio of long and short term debt (items dltt and dlc) to the sum of long and short term debt plus common equity (items dltt, dlc and ceq) and Tobin s q as the ratio of firm s total market value (item prcc f times the absolute value of item csho plus items at and ceq minus item txdb) over total assets (item at). CAPX is total capital expenditures (item capx) over total assets (item at). Accruals are the discretionary accruals calculated using the modified Jones model as in Dechow et al. (1995). Market Cap is the firm s total market value (item prcc f times the absolute value of item csho plus items at and ceq minus item txdb). All variables are winsorized at the 1 percent level. As commonly done, we exclude financial, utilities and governmental and quasi governmental firms (SIC codes from 6000 to 6999, from 4900 to 4999 and bigger than 9000; respectively) both because their measure of return on assets may not be appropriate and/or because their competition for managerial talent may be distorted. We use the 49 Fama-French Industry classification: our final sample includes 36 different industries. Our principal measure of firm corporate governance is the Gompers et al. (2003) governance index, which we obtain from RiskMetrics. The G-Index ranges from 1 to 24 and one point is added for each governance provision restricting shareholders right with respect to managers (for further details see Gompers et al. 2003). 7 A higher G-Index indicates more restrictions on shareholder rights or a greater number of anti-takeover measures. Therefore, a higher value of the G-Index corresponds to a lower g in our theoretical representations. Hence, all coefficient signs on the 7 The list of provisions included in the G-Index are as follows: Antigreenmail, Blank Check, Business Combination laws, Bylaw and Charter amendment limitations, Control-share Cash-out laws, Classified Board (or Staggered Board), Compensation Plans, Director indemnification Contracts, Control-share Acquisition laws, Cumulative Voting, Directors Duties provisions, Fair-Price provisions, Golden Parachutes, Director Indemnification, Limitations on director Liability, Pension Parachutes, Poison Pills, Secret Ballot, Executive Severance agreements, Silver Parachutes, Special Meeting limitations, Supermajority requirements, Unequal Voting rights, and Limitations on action by Written Consent. 19

21 empirical predictions using the G-Index switch sign with respect to the ones using our theoretical g governance measure. To fill the gaps between reported values, we follow Gompers et al. (2003) and assume that any change happens at the end of the missing period. 8 We obtain our measures of executive compensation from ExecuComp focusing on the CEO as the manager. We measure Total Compensation as natural logarithm of item tdc1. We define Pay for Performance as the ratio of bonuses and stock options (the latter is the natural logarithm of the Black Scholes value of options granted: item option awards blk value) and total compensation, measured in percentage terms. We also use ExecuComp to define: CEO Tenure as the difference between the current year and the year the executive became CEO (item becameceo); CEO Age as the age of the CEO and External as a dummy variable that takes value one if the CEO was not an executive in the firm the year before being appointed as CEO, and zero otherwise. We also control for CEO Duality, which is a dummy variable that takes the value 1 if the CEO is also the Chairman of the board. We obtain data on takeovers from Thompson Deals. We require the percentage of shares held by the acquiror after the transaction to be higher than 20% and the percentage of shares held by the acquiror 6 months prior to the transaction to be lower than 20%. We observe 462 takeovers in our merged sample. We define Takeover as a dummy variable that takes value 1 if that manager-firm match experiences a takeover, zero otherwise. Takeover Completed takes value 1 if the variable Takeover takes value one and the deal was ultimately completed. Summary statistics for all the variables are reported in Table 1. Our dataset spans the period from 1993 to 2007 as this corresponds to the RiskMetrics data availability. Moreover, Appendix B includes details on the individual components of the G-Index. 4.4 Results To show that indeed firms choose weaker governance to attract better quality managers, first we need to estimate CEO fixed effects. In Table 2, we show the results 8 We check for robustness by using linear interpolation, finding no significant change in the results. 20

22 from regression (7) with several time dependent regressors (X it ) and time independent industry fixed effects (z ind ). We report the regression coefficients, overall fit of the model and some descriptive statistics of the estimated CEO fixed effects. We report the mean, minimum, maximum and standard deviation of the estimated CEO ability to show that CEO choice does indeed matter for firm performance Cross-sectional evidence Table 3 presents our main empirical result. In Panel A we find empirical support for equation (9); we show that better managers are employed by firms with lower corporate governance (Column 1), are paid more (Columns 2 and 3) and work in bigger firms (Columns 4 and 5). We use Weighted-Least-Squares estimators, where the weights are the inverse of the standard deviation of the CEO fixed effects estimated in the first stage. We control for industry/year fixed effects, and CEO characteristics (CEO tenure, age and external dummy). In Column 1 we consider the relation between corporate governance and managerial ability. As predicted by the model, increases in managerial quality are associated with decreases in governance. In Column 2 and 3, we report the correlations between managerial talent (as proxied by the CEO fixed effect) and total compensation and pay for performance. Better managers are paid more, and are paid more in the form of flexible pay (bonuses and options). In Column 4, we also confirm that better CEOs work in larger firms, as also argued by Gabaix and Landier (2008). In Column 5, we show that better managers also invest more (in terms of capital expenditure), an alternative interpretation of the parameter I in our model. In terms of economic magnitude, Panel A implies that holding all else constant, one standard deviation increase in CEO talent (which corresponds to an increase by according to Table 2) implies a 0.4 point increase in G-Index (or decrease in governance), and a 14% increase in flexible pay. In Panel B, we test the model predictions regarding takeover and CEO employment lengths. Columns (1) and (2) find empirical support for equation (8) with takeover as the dependent variable. Column (1) analyzes all deals while column (2) focuses only on completed transactions. Moreover, better-quality managers stay 21

23 longer in their firms. Using the estimated measure of CEO quality, we test a Cox model (in Columns 3 and 4). Column 3 present our baseline analysis, while column 4 focuses on those CEOs under 65 years of age as these CEOs are less likely to be affected by retirement. Overall, we find support for our hypothesis in all the specifications: one standard deviation increase in CEO ability leads to around 14% decrease in the hazard rate (column 3). In un-tabulated results, we estimate the model with constant hazard rates and find very similar results Time-series evidence In Table 4, we sharpen the test of the key prediction of the model by looking at changes in governance around CEO turnovers. If poor corporate governance is chosen as part of the CEO incentive contract to attract better quality managers, we would expect that changes in corporate governance should be more common in times when the CEO is turned over. Moreover, we would expect governance to increase when the new CEO is of lower quality than the earlier CEO; and vice-versa, governance should decrease when the new CEO is of better quality than the older one. Changes in governance happen in 31 percent of the observations: in 22 percent of the cases governance worsens (as the G-Index increases) while in 9 percent of the cases governance improves (as the G-Index decreases). There is a CEO turnover in about 19 percent of the observations. In 8 percent of the observations, the new CEO is of better quality than the earlier one (Turnover Up), while in 10 percent of the cases the new CEO is of worse quality of the earlier one (Turnover Down). In Panel A of Table 4, we conduct the main test. In Column 1, we show that CEO turnovers are associated with a higher frequency of governance change. This is consistent with the model s assumption that governance is chosen as part of the CEO incentive scheme. In Columns 2 and 3, we test whether governance increases around turnover when the new CEO is worse than the old one and decreases when the new CEO is better than the old one. The indicator Turnover Up is indeed positively correlated with increases in governance in Column 2, highlighting the role of lower corporate governance as part of the optimal compensation contract to attract high quality managers. Column 3 reports however that the employment of worse managers 22